On 2/12/24 22:38, Suren Baghdasaryan wrote:
> Currently slab pages can store only vectors of obj_cgroup pointers in
> page->memcg_data. Introduce slabobj_ext structure to allow more data
> to be stored for each slab object. Wrap obj_cgroup into slabobj_ext
> to support current functionality while allowing to extend slabobj_ext
> in the future.
>
> Signed-off-by: Suren Baghdasaryan <surenb@xxxxxxxxxx>
...
> +static inline bool need_slab_obj_ext(void)
> +{
> + /*
> + * CONFIG_MEMCG_KMEM creates vector of obj_cgroup objects conditionally
> + * inside memcg_slab_post_alloc_hook. No other users for now.
> + */
> + return false;
> +}
> +
> +static inline struct slabobj_ext *
> +prepare_slab_obj_exts_hook(struct kmem_cache *s, gfp_t flags, void *p)
> +{
> + struct slab *slab;
> +
> + if (!p)
> + return NULL;
> +
> + if (!need_slab_obj_ext())
> + return NULL;
> +
> + slab = virt_to_slab(p);
> + if (!slab_obj_exts(slab) &&
> + WARN(alloc_slab_obj_exts(slab, s, flags, false),
> + "%s, %s: Failed to create slab extension vector!\n",
> + __func__, s->name))
> + return NULL;
> +
> + return slab_obj_exts(slab) + obj_to_index(s, slab, p);
This is called in slab_post_alloc_hook() and the result stored to obj_exts
but unused. Maybe introduce this only in a later patch where it becomes
relevant?
> --- a/mm/slab_common.c
> +++ b/mm/slab_common.c
> @@ -201,6 +201,54 @@ struct kmem_cache *find_mergeable(unsigned int size, unsigned int align,
> return NULL;
> }
>
> +#ifdef CONFIG_SLAB_OBJ_EXT
> +/*
> + * The allocated objcg pointers array is not accounted directly.
> + * Moreover, it should not come from DMA buffer and is not readily
> + * reclaimable. So those GFP bits should be masked off.
> + */
> +#define OBJCGS_CLEAR_MASK (__GFP_DMA | __GFP_RECLAIMABLE | \
> + __GFP_ACCOUNT | __GFP_NOFAIL)
> +
> +int alloc_slab_obj_exts(struct slab *slab, struct kmem_cache *s,
> + gfp_t gfp, bool new_slab)
Since you're moving this function between files anyway, could you please
instead move it to mm/slub.c. I expect we'll eventually (maybe even soon)
move the rest of performance sensitive kmemcg hooks there as well to make
inlining possible.
> +{
> + unsigned int objects = objs_per_slab(s, slab);
> + unsigned long obj_exts;
> + void *vec;
> +
> + gfp &= ~OBJCGS_CLEAR_MASK;
> + vec = kcalloc_node(objects, sizeof(struct slabobj_ext), gfp,
> + slab_nid(slab));
> + if (!vec)
> + return -ENOMEM;
> +
> + obj_exts = (unsigned long)vec;
> +#ifdef CONFIG_MEMCG
> + obj_exts |= MEMCG_DATA_OBJEXTS;
> +#endif
> + if (new_slab) {
> + /*
> + * If the slab is brand new and nobody can yet access its
> + * obj_exts, no synchronization is required and obj_exts can
> + * be simply assigned.
> + */
> + slab->obj_exts = obj_exts;
> + } else if (cmpxchg(&slab->obj_exts, 0, obj_exts)) {
> + /*
> + * If the slab is already in use, somebody can allocate and
> + * assign slabobj_exts in parallel. In this case the existing
> + * objcg vector should be reused.
> + */
> + kfree(vec);
> + return 0;
> + }
> +
> + kmemleak_not_leak(vec);
> + return 0;
> +}
> +#endif /* CONFIG_SLAB_OBJ_EXT */
> +
> static struct kmem_cache *create_cache(const char *name,
> unsigned int object_size, unsigned int align,
> slab_flags_t flags, unsigned int useroffset,
> diff --git a/mm/slub.c b/mm/slub.c
> index 2ef88bbf56a3..1eb1050814aa 100644
> --- a/mm/slub.c
> +++ b/mm/slub.c
> @@ -683,10 +683,10 @@ static inline bool __slab_update_freelist(struct kmem_cache *s, struct slab *sla
>
> if (s->flags & __CMPXCHG_DOUBLE) {
> ret = __update_freelist_fast(slab, freelist_old, counters_old,
> - freelist_new, counters_new);
> + freelist_new, counters_new);
> } else {
> ret = __update_freelist_slow(slab, freelist_old, counters_old,
> - freelist_new, counters_new);
> + freelist_new, counters_new);
> }
> if (likely(ret))
> return true;
> @@ -710,13 +710,13 @@ static inline bool slab_update_freelist(struct kmem_cache *s, struct slab *slab,
>
> if (s->flags & __CMPXCHG_DOUBLE) {
> ret = __update_freelist_fast(slab, freelist_old, counters_old,
> - freelist_new, counters_new);
> + freelist_new, counters_new);
> } else {
> unsigned long flags;
>
> local_irq_save(flags);
> ret = __update_freelist_slow(slab, freelist_old, counters_old,
> - freelist_new, counters_new);
> + freelist_new, counters_new);
I can see the mixing of tabs and spaces is wrong but perhaps not fix it as
part of the series?
> local_irq_restore(flags);
> }
> if (likely(ret))
> @@ -1881,13 +1881,25 @@ static inline enum node_stat_item cache_vmstat_idx(struct kmem_cache *s)
> NR_SLAB_RECLAIMABLE_B : NR_SLAB_UNRECLAIMABLE_B;
> }
>
> -#ifdef CONFIG_MEMCG_KMEM
> -static inline void memcg_free_slab_cgroups(struct slab *slab)
> +#ifdef CONFIG_SLAB_OBJ_EXT
> +static inline void free_slab_obj_exts(struct slab *slab)
Right, freeing is already here, so makes sense put the allocation here as well.
> @@ -3817,6 +3820,7 @@ void slab_post_alloc_hook(struct kmem_cache *s, struct obj_cgroup *objcg,
> kmemleak_alloc_recursive(p[i], s->object_size, 1,
> s->flags, init_flags);
> kmsan_slab_alloc(s, p[i], init_flags);
> + obj_exts = prepare_slab_obj_exts_hook(s, flags, p[i]);
Yeah here's the hook used. Doesn't it generate a compiler warning? Maybe at
least postpone the call until the result is further used.
> }
>
> memcg_slab_post_alloc_hook(s, objcg, flags, size, p);
